DESCRIPTION: Understanding the underlying mechanisms of the brain and behavior is an essential requirement for improving the diagnosis of and treatments for mental health diseases. There is an intense interest in generating different modulates of neuroimaging and behavioral data to gain new insights into brain functionality and its connections with behavior outcomes. Challenges in studying the relationships, particularly the alignment, among brain- behavior outcomes from different sources include: (1) the outcomes measured from different modalities are diverse and complex, often in different scales (continuous, ordinal) and of different data representations(scalar, vector, matrix); (2) the neuroimaging data is high dimensional and reflects not only the desired signal but also the background noise; (3) there are obstacles to compare neuroimaging data from multi-center studies due to considerable between-center variability in brain images obtained from different scanners and processing protocols. Currently, there are very limited statistical methods available to address these issues. The overall objective of this proposal is to develop a unified statistical framework that fills in the aforementioned gaps. Our proposal of adopting agreement-based methodology provides a novel perspective for investigating the alignment between behavior outcomes and the biology of the brain (neuroimaging). While standard agreement methodology has been limited to the evaluation of outcomes that are made on the same scale, our seminal work on broad sense agreement (BSA) (Peng et al. 2011, a featured JASA article) that characterizes the agreement/alignment between a continuous and an ordinal variable lays the foundation for a promising framework proposed in this application. Specifically, we plan to fulfill our research goals by characterizing the alignment among outcomes with different scales and data-representations; incorporating covariates; assessing the strength of alignment between neuroimaging biomarkers and symptom domains; identifying relevant features in high-dimensional neuroimaging data that align with specific symptom clusters; and assessing agreement and calibrating images from multi-center studies. The proposed statistical methods will be applied to an ongoing PTSD study and a national multi-center imaging study, and user-friendly software will be developed and made available to general research communities. Our proposed method developments will directly benefit mental health research, and they are ubiquitous enough to be generally useful contributions to statistical practice.